[U-Boot] RAM Probing for ads5121rev4
Hi all, In your opinion it is possible to probe the ram before initialize it? Unfortunately there are some rev4 which are equipped with different ram model. Examples are welcomed.
B.R. Emanuele
Dear Emanuele Placidi,
In message 7c4b93310902180021o56da8320w2084e6a8c09c9509@mail.gmail.com you wrote:
In your opinion it is possible to probe the ram before initialize it? Unfortunately there are some rev4 which are equipped with different ram model. Examples are welcomed.
It is a standard feature in U-Boot (and has always been right from the start with PPCBoot nearly 10 years ago) that we auto-detected the memory configuration on the boards, both for RAM and flash. Some boards / architectures did not bother to implement this, but the feature itself is present.
The details depend on processor, memory controller properties, specific RAM types used, etc.
In case of the Micron/Elpida issue on the ADS5121, that should be not too difficult to solve.
Best regards,
Wolfgang Denk
Wolfgang Denk wrote:
Dear Emanuele Placidi,
In message 7c4b93310902180021o56da8320w2084e6a8c09c9509@mail.gmail.com you wrote:
In your opinion it is possible to probe the ram before initialize it? Unfortunately there are some rev4 which are equipped with different ram model. Examples are welcomed.
It is a standard feature in U-Boot (and has always been right from the start with PPCBoot nearly 10 years ago) that we auto-detected the memory configuration on the boards, both for RAM and flash. Some boards / architectures did not bother to implement this, but the feature itself is present.
The details depend on processor, memory controller properties, specific RAM types used, etc.
In case of the Micron/Elpida issue on the ADS5121, that should be not too difficult to solve.
The issue concerns the possibility to identify DDR chips type (e.g. by VendorId) before initialize them. It doesn't concern configuration parameters (size, bus frequency, ecc..). There isn't a DDR command to read this information, do you think it is possible to solve this problem in other ways?
Best regards,
Wolfgang Denk
Dear "Michele De Candia (VT)",
In message 49D23E30.4060205@valueteam.com you wrote:
The issue concerns the possibility to identify DDR chips type (e.g. by VendorId) before initialize them. It doesn't concern configuration parameters (size, bus frequency, ecc..). There isn't a DDR command to read this information, do you think it is possible to solve this problem in other ways?
Yes, of course it is possible to solve this problem otherwise.
In a sane hardware design you can read this information from some board configuration register or so; for example, on the ADS5121 there could be such a register implemented in the CPLD. But alas, there ain't none.
Martha Marx posted pretty good patches (identifyung board revisions based on the MAC address) more than a month ago, but then failed to address the last review comments.
If you want to get that fixed, please pick up her patches (or nudge her to do it for you).
Best regards,
Wolfgang Denk
Wolfgang Denk wrote:
Dear "Michele De Candia (VT)",
In message 49D23E30.4060205@valueteam.com you wrote:
The issue concerns the possibility to identify DDR chips type (e.g. by VendorId) before initialize them. It doesn't concern configuration parameters (size, bus frequency, ecc..). There isn't a DDR command to read this information, do you think it is possible to solve this problem in other ways?
Yes, of course it is possible to solve this problem otherwise.
In a sane hardware design you can read this information from some board configuration register or so; for example, on the ADS5121 there could be such a register implemented in the CPLD. But alas, there ain't none.
Martha Marx posted pretty good patches (identifyung board revisions based on the MAC address) more than a month ago, but then failed to address the last review comments.
If you want to get that fixed, please pick up her patches (or nudge her to do it for you).
Hi Mr. Denk, thanks for the suggestions, we've already picked up Martha Marx patches and they work fine on ADS5121. We want to avoid DDR identification by board revisions because we don't know which DDR will be used in the future and it isn't a flexible way to do that. After all, if it is the only way, we'll use this one.
Thanks, Michele De Candia
Best regards,
Wolfgang Denk
participants (3)
-
Emanuele Placidi -
Michele De Candia (VT) -
Wolfgang Denk